AI Implementation · Ontario SMBs

AI that actually ships
into your workflow.

I build AI-powered integrations for Ontario SMBs — document extraction pipelines, email triage agents, internal knowledge search, automated reporting. Working systems in weeks, built into your existing stack, no vendor lock-in.

Book a free 30-minute AI fit call Not sure AI fits your business? That's literally what the call is for.
2–4 week sprints
Vendor-neutral
Model-agnostic
Ontario SMBs
Models we work with
Claude Sonnet 4.5 GPT-4o Gemini 2.0 Flash Azure OpenAI + local & open-source
The Gap

Most AI projects die between the demo and the deployment.

You've seen what AI can do. Maybe your team already has Claude, Copilot, or a vendor AI subscription. But between an impressive demo and a workflow that actually runs reliably — with the right model, a prompt pipeline that holds up on real inputs, and outputs your team trusts — there's a serious engineering gap.

Writing reliable prompt pipelines, picking the right model for the task, wiring it into your email or CRM or document system, handling edge cases: that's the work. Most businesses under 200 people don't have anyone whose job it is to do it.

That's the job I do.

How It Works

From first call to working system.

Step 01
Fit Call + Assessment
We spend 30 minutes on your highest-friction workflow. If there's a real AI opportunity, I scope it in writing: what gets built, what model fits, what it costs. No obligation.
Week 0–1 · Free
Step 02
Build Sprint
Fixed scope, fixed fee. I build the integration in your actual environment — connected to your email, your docs, your tools. Weekly check-ins so there are no surprises.
Weeks 2–4 · Fixed fee
Step 03
Hand-off & Own It
You get working code, a plain-English handoff doc, and a walkthrough call. Your team runs it — no dependency on me. Optional retainer if you want ongoing model updates or new sprints.
Week 4–5 · You own the code
How We Work

Three ways to work together.

AI Readiness Assessment
$500–$1,500 · Fixed fee
A focused audit of your workflows, data, and tools. You get a written report with the top 3 AI opportunities in your business, ranked by ROI and effort — plus an honest "not worth it" list for the things AI can't do well yet.
Timeline: 1 week
Written report + 1-hour walkthrough call
Most popular
Integration Sprint
$2,000–$8,000 · Fixed scope, scoped on the fit call
One specific AI use case, implemented end-to-end in your real environment. Built on the right model for your stack — Claude, GPT-4o, or Gemini — and integrated into your existing tools: Gmail, Outlook, SharePoint, Notion, or a custom internal system. Deliverables: working code, a handoff doc, and a walkthrough so your team can own it.
Timeline: 2–4 weeks
Working system + docs + handoff training
Fractional AI Lead
$1,500–$3,000/month · Cancel anytime
Ongoing access for model evaluation (Claude, GPT-4o, Gemini — picking the right tool per job), new use case scoping, and keeping your systems current as the model landscape shifts fast. The senior AI engineer you can't justify full-time, at a fraction of the cost.
Ongoing · quarterly review
Monthly roadmap + priority support

Also available: half-day Team Enablement Workshops ($500–$1,000) — hands-on training so your non-technical team uses AI tools safely and effectively. First two clients receive a 50% case-study rate in exchange for a testimonial and real before/after numbers.

Why Praxis

What you're actually getting.

Vendor-neutral
I recommend whichever model fits your use case — Claude Sonnet for reasoning-heavy tasks, GPT-4o for multimodal, Gemini for Google Workspace — or an off-the-shelf tool you already own. You're not locked into my stack or anyone else's.
Research-grade, not buzzword-grade
Production experience with Claude, GPT-4o, Gemini, RAG pipelines, MCP integrations, and multi-step agentic systems — plus the research background to cut through vendor hype. I read the papers so you don't have to.
Weeks, not quarters
Scoped engagements with a defined deliverable and end date. No open-ended billable-hours arrangements. You know what you're getting and when.
Honest about limits
Part of every assessment is the list of things AI shouldn't do in your business yet. You'll never get a recommendation I can't defend technically.
Built in Public

See it before you buy it.

No client logos yet — here's something better: working demos built on real AI models with synthetic data. Each solves a real problem for the industries I work with. Watch the 90-second walkthroughs, or see it live on a fit call.

Manufacturing
RFQ Email → Structured Quote Extractor
Reads messy RFQ emails, extracts specs, quantities and deadlines into schema-locked fields, prices them against the shop's price book, and drafts the reply — with follow-up questions for anything missing. Your estimator reviews instead of re-keys.
See it live — book a fit call
Property Management
Email Triage + Auto-Draft Response Agent
Triages the whole inbox in one pass: classifies tenant, owner and vendor email by type and urgency, extracts building and unit, drafts first-pass replies — and flags anything with legal or money exposure for human judgment first.
See it live — book a fit call
Professional Services
Internal Document Search with Citations
Ask plain-language questions over internal SOPs, policies and spec documents. Every answer cites the exact source section — and when the answer isn't in your documents, it says so instead of guessing.
See it live — book a fit call
Common Questions

Things people ask before booking.

Do we need to have AI experience in-house?
No. That's the point. You know your business and your bottlenecks — I handle the model selection, prompt engineering, and integration work. Your team learns what they need to run it on the handoff call.
We already use Microsoft Copilot / ChatGPT. Does that conflict?
Not at all — I'll either work with what you have or explain honestly if a different model would do the specific job better. The goal is the right tool for the task, not replacing your existing subscriptions.
How do you pick which AI model to use?
It depends on the task: Claude Sonnet for multi-step reasoning and document work, GPT-4o when vision or Microsoft integrations are involved, Gemini for Google Workspace-heavy stacks. I'll tell you the tradeoffs before anything is built.
Can you work with our existing software and data?
Yes — integrating into your existing stack (not replacing it) is the whole model. Most projects connect to Gmail, Outlook, SharePoint, Notion, a CRM, or internal databases. If you're unsure, the fit call is the place to find out.

Find out if AI is worth it for your business — in 30 minutes, for free.

No deck, no pitch. Bring one workflow that eats your team's time. I'll tell you on the call whether AI can fix it, roughly what it would take, and what it would cost. If the answer is "it can't," you'll know that too.

Book the fit call